Sample Sizing Approaches for Container Closure Integrity (CCI) Testing

Introduction

Sample size approaches for container closure integrity testing (CCIT) for routine commercial biopharmaceutical production vary across the industry. Currently, there is no official standard or published procedure to follow for sampling approaches or defining appropriate sample sizes.

Unpublished industry surveys highlight that the majority of leading biopharmaceutical companies do not routinely perform CCIT for each batch of a commercial product but rather rely on a holistic approach guaranteeing consistent quality of the finished parenteral product with respect to CCI (container closure integrity).¹ CCIT is not a batch release requirement for the majority of companies; some companies may sample lots for CCIT as part of in-process control (IPC) or as a characterization test to monitor performance of the entire manufacturing process. This is consistent with current regulatory guidance, which does not require CCIT for every container, or for every batch, unless the container is sealed by fusion,² which is out of the scope of this paper. CCIT in support of container closure system (CCS) development, process validation activities or stability testing are also out of scope.

This industry group considers that a holistic approach^3,4 is preferable to sampling and testing for assuring CCI. However, it is accepted and appropriate that, when sampling is required, scientifically valid sampling plans should be based on risk assessment, including information such as supplier approval, packaging component specifications and process knowledge

This paper provides some considerations and guidance on defining a justifiable sample size for CCIT when required for parenteral biopharmaceutical finished products. This is a particularly important consideration when using destructive test methods.

This paper summarizes a practical and risk-based strategy, both representing scientifically justified approaches. These approaches prevent unnecessary sampling and testing while retaining high product quality. Where the integrity of each individual container is assured by a suitably qualified process, additional CCI sampling and testing would not normally be required.

It is not the intention of this paper to prescribe an approach that represents the best fi t for all biopharmaceutical companies or product presentations. Rather, the aim is to illustrate some concepts that facilitate the development of scientifically justified sampling plans that need not represent an unnecessary burden on industry.

Frequently, it is perceived that an increased sample size is necessary for larger batch sizes. Indeed, commonly applied sampling plans such as ANSI Z1.4⁴ and MIL-Std-105E⁵ specify larger sample sizes for larger batch sizes. This paper proposes that sampling, based on the level of defects, used in conjunction with these commonly applied sampling plans gives a more appropriate approach.

Prerequisites

Biopharmaceutical manufacturers are presumed to have a robust holistic control strategy to support CCI for parenteral finished drug products. The discussions of sampling approaches and number of samples described in this paper only apply if a robust control strategy is already in place.

A robust CCI control strategy should include, but is not limited to:

• Robust design of the container closure system (CCS), including the qualification status of the suppliers of the CCS components
• Established dimensional control strategy for components
• Validated manufacturing process including CCI related parameters
• Appropriate in-process controls (IPC) during manufacturing
• Visual inspection of each finished drug product container with good detectability of visible CCI-relevant defects such as breakage, plunger defects, and seal defects

It is the responsibility of each individual biopharmaceutical manufacturer to ensure that it develops an appropriate control strategy and is able to defend its own strategy supported by appropriate data, process/component controls and manufacturing experience.

The sampling approaches discussed in this paper are based on treating CCI test results as attribute data (i.e., pass/fail) since most CCIT methods process the test results as such.

Common Sampling Approach

From a historical and practical perspective, CCIT sampling has often been based on the requirements to demonstrate sterility of parenteral product lots (e.g., USP<71>, Ph. Eur. 2.6.1).

Typically, a zero-acceptance sampling plan may be used, where a total of 20 finished product samples are considered as a minimum sample size also for routine testing of CCIT. In some cases, a sample of this size cannot adequately represent the population (e.g. integrity of all units in a batch) in terms of the defect rate. For example, in Table 1, where the frequency of defects was 0.001 (1 in a 1000), testing of 20 units from a lot of any size, would represent a probability of a failing result (sterility or CCIT) of 0.0198 (or 2%). Alignment of sample size between sterility and CCIT is justifiable and appropriate as CCIT addresses a related quality attribute.

This sample size is only justifiable if the product quality (e.g., CCI and sterility) is intentionally designed into the product during processing and not just “guaranteed” by final product testing. It is therefore presumed that the product has been manufactured under conditions designed to exclude contamination and to provide consistent quality of the closure (i.e., by applying a holistic approach to CCI).

Risk-Based Sampling Plan Approach

An adequate sampling plan gives assurance that is appropriate to the criticality of the defect being inspected. While CCI is generally considered a critical defect based on patient safety risk, each company should develop its own risk assessment procedure for CCI. The procedure should defi ne the maximum allowable defect level for each defect category.

Based on these criteria, along with other data obtained during process characterization and/or container development phases, a scientifically based decision can be made to determine the appropriate acceptance quality level (AQL) and limiting quality (LQ^b) level for CCI.

An effective risk-based sampling plan shall have a high probability of accepting a lot that is considered to be of good quality and a low probability of accepting a lot that is considered to be of poor quality. The following sampling plans are equally justifiable for CCI testing and may present an alternative to the common sample size approach if desired

ANSI/ASQ Z1.4, ISO2859-1, or Zero Acceptance Number Sampling Plan

ANSI/ASQ Z1.4, ISO 2859-1, and Zero Acceptance Number sampling plan (or C=0 Sampling Plan)^7,8,9 are the mainstream standards/methods commonly used by the companies for the purpose of lot acceptance or lot rejection. ANSI/ASQ Z1.4 and ISO 2859-1 were both adapted from the military standard MIL-STD-105D but with different normal/tightened/reduced switching rules.

Zero Acceptance Number sampling plan is derived from ANSI/ASQ Z1.4 and is based on the premise of accepting a lot only when zero defects are found during inspection. Zero Acceptance Number sampling plan requires a smaller sample size and is usually used when the company has high confidence in the quality of the attribute being inspected and therefore willing to assume a lower risk of rejecting a good lot. These standards focus on producer’s risk and what the plan routinely accepts but not necessarily the product quality. By specifying the inspection level, AQL and lot size, sampling plans for CCI can be easily determined using these standards.

ANSI/ASQ Z1.4 and ISO 2859-1 contain three inspection levels (I, II, III) for general use. When relatively small sample sizes are necessary or large sampling risks can or must be tolerated, four inspection levels (S-1, S-2, S-3, S-4) for special use can be applied.^c

Assuming a robust control strategy is in place for CCI and the process is validated, it is acceptable to use the inspection level for special use when testing the CCI. Special inspection level S-4 may be a good choice to use since it gives the most discriminating sample size among the four special inspection levels.

Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special offers
from American Pharmaceutical Review – all delivered right to your inbox! Sign up now!

Table 2 presents a different sampling scheme using the reduced S-4 special inspection level and the Single sampling plan. This is assuming that by applying the ANSI switching rule, reduced inspection is in effect as the conditions are met during the normal inspection period in a continuous series of lots. In this example (Table 2), if a manufacturer produces a lot with 20,000 units, a total of 20 units would be pulled from the lot for CCIT. All units must comply with the CCIT acceptance criteria to accept the lot. The AQL (0.25% in this example) reflects the worst tolerable average defect rate for the process and will pass inspection with high probability (e.g., 95%). The AQL is a parameter of the sampling scheme – it is not the process average. However, this sampling scheme does not ensure detection of low frequency defects in a single lot. For example, if a batch had a true defect level of 1% and 20 units were sampled for CCIT, there would still be an 82% probability that all units would pass and the lot would be accepted.

It is important to note that when sampling, the units should be sampled in a manner that is representative of the whole lot.

Operating Characteristic Curves

Operating Characteristic (OC) Curves can be constructed to quantify sampling risk graphically.

For example, when a manufacturer samples 20 units for CCIT and only accepts the lot when zero defects are found, they can be 95% confident that the lot defect rate is equal to or less than 13.91%. There is 95% probability of accepting a lot with 0.26% defects.

If a batch had a true defect level of 1% and 20 units were sampled for CCIT, there would still be an 82% probability that all units would pass and the lot considered to be of poor quality would be accepted.

Examples of sampling plans that give different levels of assurance with their corresponding limiting quality are provided in Table 3. The example sample sizes presented here were calculated based on accepting lots only when zero defects are found. Other acceptance/rejection numbers are acceptable as long as the associated AQL/LQ meet each company’s requirement based on their risk assessment.

Conclusion

Container closure integrity testing may be conducted on selected lots to confirm the performance of a robust holistic control strategy at regular intervals or in case of a market-specific regulatory requirement. Both approaches, commonly used and risk-based statistical, are appropriate in defining the CCIT sample size for testing of a lot.

The sampling approaches discussed are justifiable if the product quality (e.g., CCI and sterility) is designed into the product and not just “guaranteed” by final product testing.

Container Closure Integrity Testing of the final product does not provide certainty that a process is well controlled. A holistic approach to CCI, however, builds knowledge and confidence around a product through process understanding and control in the spirit of concepts like quality by design. When critical-to-quality parameters are assessed and controlled during the development and qualification phases, with on-going monitoring, issues can be averted, or mitigated.¹²

In consideration of a holistic approach (see prerequisites) and presuming a stable and well-controlled manufacturing process, a sampling approach for CCIT does not provide additional assurance for the integrity of each manufactured unit within a lot, nor does it confer additional quality on the product.

Some sampling and testing of products for container closure integrity may be periodically required. This paper outlines approaches to sampling which are scientifically justifiable, yet also demonstrate that a statistically-relevant sampling plan need not require an unnecessary level of testing, particularly with reference to destructive testing. The strategies described here are not exhaustive but serve only to illustrate possible approaches that might be applied by a biopharmaceutical company if required to sample and test drug product lots for CCI.

About BioPhorum

BioPhorum comprises over 90 manufacturers and suppliers deploying their top 3,500 leaders and subject matter experts to work in seven focused Phorums, articulating the industry’s technology roadmap, defining the supply partner practices of the future, and developing and adopting best practices in drug substance, fill finish, process development and manufacturing IT. In each of these Phorums, BioPhorum facilitators bring leaders together to create future visions, mobilize teams of experts on the opportunities, create partnerships that enable change and provide the quickest route to implementation, so that the industry shares, learns and builds the best solutions together.

Disclaimer

This document represents a consensus view, and as such does not represent the internal policies of the contributing companies. Neither BioPhorum nor any of the contributing companies accept any liability to any person arising from their use of this document. The views and opinions contained herein are that of the individual authors/contributors and should not be attributed to the employers of such authors/contributors.

References

1. Ewan S, Jiang M, Stevenson C, et al. White Paper: Container closure integrity control versus integrity testing during routine manufacturing, PDA J Pharm Sci and Tech 2015, 69 461-
2. EudraLex, 2008, The Rules Governing Medicinal Products in the European Union, Volume 4, EU Guidelines to Good Manufacturing Practice Medicinal Products for Human and Veterinary Use; Annex 1 Manufacture of Sterile Medicinal Products
3. ECA. Container Closure Integrity testing of medicinal products for parenteral use. Position Paper
4. Brown H, Mahler HC, Mellman J et al Container Closure Integrity Testing—Practical Aspects and Approaches in the Pharmaceutical Industry. PDA Journal of Pharmaceutical Science and Technology. 2017 Mar 1;71(2):147-62.
5. ANSI Z1.4-2003, Sampling Procedures and Tables for Inspection by Attributes
6. MIL-STD-1916, “DoD Preferred Methods for Acceptance of Product
7. Rapid Sterility Testing”, Sutton, S. 2011. Sterility Tests IN Rapid Sterility Testing J. Moldenhauer (ed) PDA/DHI Publ pp 7-24 ISBN: 1-933722-56-8.
8. ANSI Z1.4-2003, Sampling Procedures and Tables for Inspection by Attributes
9. ISO 2859-1:1989, Sampling procedures for inspection by attribute – Part 1: Sampling plans indexed by acceptable quality level (AQL) for lot-by-lot inspection
10. Squeglia, N. L., Zero Acceptance Number Sampling Plans, Fifth Edition, ASQ Press, ISBN 978-0-87389-739-6
11. Use of F distribution for estimating AQL and LQL - page 104 of Hahn, G. D. and Meeker, W.Q., 1991, Statistical Intervals: A Guide for Practitioners, John Wiley & Sons, Inc., New York, 392 pp.
12. Use of Binomial distribution for calculating Pr (Accept) - page 614 of Montgomery, D.C., 1997, Introduction to Statistical Quality Control, 3rd Edition, John Wiley & Sons, Inc., New York, 677 pp.

Non-Cited references

• Hanson, J., Determination of Sampling Plans Utilizing Risk Analysis, https://web.archive.org/web/20190214201211/http://www.cbinet.com/sites/default/files/files/Hanson_Jeff_Pres-UPDATED.pdf
• Varney, M., Statistically Based Validation Acceptance Criteria, http://www.mbswonline.com/upload/presentation_5-24-2012-10-31-44.ppt

1. Note: This table employs a poisson distribution as described in reference 6, rather than a binomial distribution which is used later in the paper
2. Limiting quality (LQ) is also called Rejectable Quality Level (RQL) and Lot Tolerance Percent Defective (LTPD)
3. As an additional note: Batch level testing, as required by some regulators, only assures us that no gross defects go undetected from the Control Strategy. This can easily be achieved with Special Sampling Plans